NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data

Abstract

Major histocompatibility complex (MHC) molecules are expressed on the cell surface, where they present peptides to T cells, which gives them a key role in the development of T-cell immune responses. MHC molecules come in two main variants: MHC Class I (MHC-I) and MHC Class II (MHC-II). MHC-I predominantly present peptides derived from intracellular proteins, whereas MHC-II predominantly presents peptides from extracellular proteins. In both cases, the binding between MHC and antigenic peptides is the most selective step in the antigen presentation pathway. Therefore, the prediction of peptide binding to MHC is a powerful utility to predict the possible specificity of a T-cell immune response. Commonly MHC binding prediction tools are trained on binding affinity or mass spectrometry-eluted ligands. Recent studies have however demonstrated how the integration of both data types can boost predictive performances. Inspired by this, we here present NetMHCpan-4.1 and NetMHCIIpan-4.0, two web servers created to predict binding between peptides and MHC-I and MHC-II, respectively. Both methods exploit tailored machine learning strategies to integrate different training data types, resulting in state-of-the-art performance and outperforming their competitors. The servers are available at http://www.cbs.dtu.dk/services/NetMHCpan-4.1/ and http://www.cbs.dtu.dk/services/NetMHCIIpan-4.0/.

Figure 1.

Example outputs for the NetMHCpan-4.1 and NetMHCIIpan-4.0 tools. (A) Example output for NetMHCpan-4.1, using as input the web server's FASTA sample data and the HLA-A*30:01 allele, with a peptide length of nine and other options set to default. (B) Example output for NetMHCIIpan-4.0, using as input the web server's FASTA sample data and the DRB1*04:34 allele, with all other options set to default. By default, prediction scores are for both methods displayed in terms of a Score_EL (the likelihood of a peptide being an MHC ligand) column and a ‘%Rank_EL’ column (the EL percentile Rank score); if the user selects to include BA predictions, such values are reported as well. The ‘BindLevel’ column displays the presence of Strong Binders (SB) or Weak Binders (WB) amongst the queried peptides. ‘Peptide’ informs the list of peptides that have been interrogated against the selected MHC molecule(s) (exhibited in the ‘MHC’ column). The ‘Pos’ entry refers to the queried peptide's position in the selected FASTA input, and ‘Core’ refers to such peptide's identified binding core. ‘Identity’ is an automatically generated ID that is assigned to the input. Other columns refer to specific properties that depend on the MHC class being employed. For additional details on the interpretation of the different columns of the output, refer to the ‘output format’ page on both web servers homepages.

Figure 2.

Epitope benchmark results for the NetMHCpan-4.1 and NetMHCIIpan-4.0 web servers. (A) Performance results for the CD8+ epitope benchmark. Median FRANK values for the different methods are: NetMHCpan-4.1, 0.00220; NetMHCpan-4.0, 0.00230; MixMHCpred, 0.00264; MHCFlurry, 0.00383; and MHCFlurry_EL, 0.00386. (B) FRANK performance results for the CD4+ epitope benchmark. The median FRANK for the different methods are: NetMHCIIpan-4.0, 0.0351; NetMHCIIpan-3.2, 0.04825; MixMH2Cpred, 0.0513; MHCnuggets, 0.1219; and DeepSeqPanII, 0.1767. (C) PPV performance results for the MS MHC class I eluted ligand benchmark. Median PPV values for the different methods are: NetMHCpan-4.1, 0.8291; NetMHCpan-4.0, 0.7940; MixMHCpred, 0.7911; MHCFlurry, 0.7256; and MHCFlurry_EL, 0.7144. P-values are shown as * P < 0.05, ** P < 10⁻⁶ and *** P < 10⁻⁹. All p-values were calculatated using a two-tailed binomial test. The plotted boxes extend from the lower to upper quartile values of the data (25^th to 75^th percentile), with a line at the median; whiskers extend from the box to show the range of the data to the most extreme, non-outlier data points.